C. elegans DAF-2 as a Model for Human InsulinReceptoropathiesDavid A. Bulger∗,‡,1, Tetsunari Fukushige∗, Sijung Yun∗, Robert K. Semple‡, John A. Hanover†,§ and Michael W. Krause∗,§∗Laboratory of Molecular Biology, †Laboratory of Molecular and Cellular Biology, National Institute of Diabetes and Digestive and Kidney Diseases, NationalInstitutes of Health, Bethesda, MD 20878, ‡Institute of Metabolic Science, University of Cambridge, Cambridge, UK, §Authors contributed equally

ABSTRACT Human exome sequencing has dramatically increased the rate of identification of disease-associated polymorphisms. However, examining the functional consequences of those variants has createdan analytic bottleneck. Insulin-like signaling in Caenorhabditis elegans has long provided a model to assessconsequences of human insulin signaling mutations, but this has not been evaluated in the context of currentgenetic tools. We have exploited strains derived from the Million Mutation Project and gene editing to explorefurther the evolutionary relationships and conservation between the human and C. elegans insulin receptors.Of 40 Million Mutation Project alleles analyzed in the C. elegans insulin-like receptor gene DAF-2, 35 exhibitedinsulin-like signaling indistinguishable from wild-type animals, indicating tolerated mutations. Five MMP allelesproved to be novel dauer-enhancing mutations, including one new allele in the previously uncharacterizedC-terminus of DAF-2. CRISPR-Cas9 genome editing was used to confirm the phenotypic consequenceof six of these DAF-2 mutations and to replicate an allelic series of known human disease mutations ina highly conserved tyrosine kinase active site residue, demonstrating the utility of C. elegans for directlymodeling human disease. Our results illustrate the challenges associated with prediction of the phenotypicconsequences of amino acid substitutions, the value of assaying mutant isoform function in vivo, and howrecently developed tools and resources afford the opportunity to expand our understanding even of highlyconserved regulatory modules such as insulin signaling. This approach may prove generally useful formodeling phenotypic consequences of candidate human pathogenic mutations in conserved signaling anddevelopmental pathways.

KEYWORDS

DAF-2INSRCRISPRMillion MutationProject

dauer

The evolutionarily conserved insulin/IGF-1 signaling (IIS) path-way is a critical regulator of growth, development, and longevity(TA N I G U C H I et al. 2006). In humans, excess insulin signalingis associated with overgrowth and cancer while loss of signal-ing results in insulin resistance and diabetes (L I N D H U R S T et al.2012). The recent expansion of human polymorphisms identifiedin genes encoding insulin signaling molecules is generating a po-tentially valuable resource for the diagnosis of metabolic disease.Full understanding of the functional consequences of these poly-morphisms ideally requires models providing in vivo context, thatis, in which the complete developmental and metabolic program

Copyright © 2016 D. A. Bulger et al.Manuscript compiled: Monday 14th November, 20161Corresponding author: Institute of Metabolic Science, University of Cambridge,Cambridge, UK. Email: dab68@cam.ac.uk

is an integral part of the phenotype.The nematode C. elegans has proven to be an excellent animal

model in which to examine insulin-like signaling. In C. elegans,reduction or partial loss of IIS has pleiotropic effects, includingaltered stress responses and increased longevity (G O T T L I E B andR U V K U N 1994). The worm IIS also controls entry into the dauerstage under adverse conditions (e.g. overcrowding, elevated tem-perature, and starvation), providing a convenient readout for thestrength of IIS signaling. Genetic screens for dauer formation de-fective (Daf-d) and dauer formation constitutive (Daf-c) mutantshave identified alleles in C. elegans genes encoding each of the IISpathway components that have helped to define and understandthis important signaling cascade (G O T T L I E B and R U V K U N 1994;M A L O N E and T H O M A S 1994; H U 2007).

Forward genetic screens in C. elegans for Daf-d or Daf-c mu-

Volume X | November 2016 | 1

INVESTIGATIONS

G3: Genes|Genomes|Genetics Early Online, published on November 16, 2016 as doi:10.1534/g3.116.037184

© The Author(s) 2013. Published by the Genetics Society of America.

tants and their suppressors impose an inherent bias due to theselection for a specific phenotype. It is unclear to what extent thatbias has limited genetic analysis of the IIS pathway. For example,the strong phenotypes selected by such forward screens may missthose alleles that more closely mimic human disease-associatedpolymorphisms that result in more subtle metabolic phenotypes.One approach that minimizes the possible bias of phenotypic selec-tion is to isolate random mutations in pathway components priorto assaying the mutant alleles for phenotypes. This approach alsoallows for the identification of silent mutations that provide infor-mation on tolerated alleles in pathway components. The MillionMutation Project (MMP) in C. elegans provides such a resource,consisting of a distributed collection of genetic alleles generatedby chemical mutagenesis and selected only for viability at roomtemperature (T H O M P S O N et al. 2013). Each MMP strain has beenclonally selected and expanded for ten generations followed bywhole genome sequencing, providing a rich array of non-lethal ho-mozygous variants within most genes of interest that are availablefor further study.

We have taken advantage of the MMP resource to further in-vestigate IIS, focusing on novel alleles of the single insulin-likereceptor in the worm, DAF-2. The DAF-2 protein has already beenextensively analyzed using the mutants obtained through geneticscreens, which resulted in a classification of the daf-2 mutants intoClass I and pleiotropic Class II based on the constellation of phe-notypes (G E M S et al. 1998; PAT E L et al. 2008). Genetic screenshave isolated several temperature-sensitive, reduction-of-functionalleles in daf-2, demonstrating that reduced IIS activity results inanimals entering the dauer stage under conditions in which wildtype animals would not. This well-defined, graded response makesDAF-2 an ideal candidate to test the value of the MMP resource inproviding additional insight into gene function. We have taken ad-vantage of a very sensitive high temperature-induced dauer assay(A I L I O N and T H O M A S 2003) to screen MMP alleles distributedthroughout the daf-2 coding region for dauer-related phenotypes.Of the 40 daf-2 MMP alleles we analyzed, 35 retained IIS functionand were indistinguishable from wild type animals, providing thefirst list of tolerated substitutions. Importantly, five MMP allelesdefined novel dauer-enhancing mutations. Using CRISPR-Cas9genome editing, we have confirmed the phenotype for many ofthese alleles and replicated an allelic series of known human dis-ease mutations in a highly conserved tyrosine kinase active siteresidue. Our results capitalize on existing MMP alleles to expandour understanding of critical residues for DAF-2 function whiledemonstrating the usefulness of C. elegans genomics to model hu-man alleles for functional studies in the context of a developingand metabolically active animal system.

MATERIALS AND METHODS

Nematode culture and strainsStandard C. elegans maintenance procedures were followed(B R E N N E R 1974). N2 (Bristol) wild-type and RW3625 let-805(st456)/qC1 [dpy-19(e1259ts) glp-1(q339)] were used as nega-tive controls and for backcrossing. The following strains contain-ing previously known daf-2 mutants were used: daf-2(e1370), JT6exp-1(sa6); daf-2(sa753), and CB120 unc-4(e120); daf-2((sa875). Thefollowing Million Mutation Project (MMP) strains containing daf-2 non-synonymous single nucleotide polymorphisms (nsSNPs)were used: VC40658 daf-2(gk748865), VC20300 daf-2(gk169914),VC40486 daf-2(gk660898), VC40818 daf-2(gk828902), VC40781 daf-2(gk808587), VC20143 daf-2(gk169915), VC20787 daf-2(gk402833),VC20470 daf-2(gk169977), VC30061 daf-2(gk407256), VC40530 daf-

2(gk680522), VC40800 daf-2(gk819989), VC30096 daf-2(gk169945),VC40395 daf-2(gk617200), VC40361 daf-2(gk595034), VC20203 daf-2(gk169944), VC40652 daf-2(gk745372), VC40958 daf-2(gk899879),VC20710 daf-2(gk379607), VC30078 daf-2(gk409799), VC40100 daf-2(gk461841), VC30098 daf-2(gk169932), VC20604 daf-2(gk351031),VC20715 daf-2(gk381236), VC30207 daf-2(gk437687), VC40848 daf-2(gk845453), VC40071 daf-2(gk452089), VC41001 daf-2(gk922267),VC40581 daf-2(gk704420), VC40041 daf-2(gk169926), VC20686 daf-2(gk373225), VC40378 daf-2(gk606297), VC40324 daf-2(gk574441),VC40329 daf-2(gk577128), VC20747 daf-2(gk390525), VC40636 daf-2(gk735782), VC40691 daf-2(gk764453), VC20785 daf-2(gk402261),VC40985 daf-2(gk913536), VC40472 daf-2(gk654266), and VC40295daf-2(gk559119). All strains were provided by the C. elegans Genet-ics Center (CGC). MMP strains were first backcrossed with malesobtained by heat-shocking the let-805/qC1 strain, and the F1 maleswere backcrossed again with let-805/qC1 hermaphrodites. To re-duce the number of background mutations in each strain, at leasttwo independent outcrossed isolates that had been isolated overlet-805 were kept for further analysis.

Strain construction

The following strains were obtained using microinjection ofthe pDD162 plasmid-based CRISPR-Cas9 system (D I C K I N S O Net al. 2013; PA I X et al. 2014) with a co-edited dpy-10(cn64)marker (A R R I B E R E et al. 2014) (Table S4): daf-2(gv51[P155L]),daf-2(gv52[S341P]), daf-2(gv53[A746T; del]), daf-2(gv54[A1015V]),daf-2(gv55[A1391T]), daf-2(gv56[A1391V]), daf-2(gv57[A1391E]), anddaf-2(gv58[A1729V]).

Sequencing of mutant alleles

Note that all INSR and DAF-2 amino acid designations refer tothe preproreceptor sequence. All daf-2 exons were sequenced us-ing nested PCR followed by Sanger sequencing with 13 ampli-cons used to cover the entire coding region (Table S1). For strainswith previously identified mutations, only the relevant exon wassequenced; MMP allele sequences were available from both theWormBase (http://www.wormbase.org/) and Million MutationProject (http://genome.sfu.ca/mmp/) websites. Strains were con-tinuously verified after genetic crosses and propagation usingSanger sequencing of the affected exon region.

Two daf-2 mutants, sa753 and sa875, isolated previously in ahigh-temperature dauer positive selection screen (A I L I O N andT H O M A S 2003) were found by whole gene sequencing of daf-2 to have the same DAF-2(P470L) substitution, which has beenpreviously reported as one of the two mutations found in mg43,DAF-2(C401Y; P470L) (K I M U R A et al. 1997).

High-temperature dauer assay

Gravid adults (1-10) were grown at 20° and allowed to lay embryosovernight (12 hours), the adults were removed and the plates wereshifted to either high temperature (27.2°) or a range of typical incu-bation temperatures (15°, 20°, and 25°) (A I L I O N and T H O M A S2003). After 72 hours at the assay temperature, the progeny werescored by counting the number of dauers, early larval stage ani-mals, and adults on each plate. Worms that had crawled up theedge of the plate were also counted (L E E et al. 2011). Some hightemperature dauer plates were left at 27.2° for another 72 hours andthen recovered at 15° for 7 days. These plates were then scored forhaving sterile adults with dead progeny (wild-type), dauers thatrecovered as fertile adults, or live dauers that remained arrested.

2 | D. A. Bulger et al.

Molecular Dynamics SimulationMolecular dynamics simulations were performed using NAMDversion 2.9 on the human INSR tyrosine kinase (PDB: 4XLV) and itspoint mutation variants. The point mutations were induced usingPyMol Molecular Graphics System, Version 1.7.6.0 Schrödinger,LLC. CHARMM Version 27 force field parameters were used togenerate solvated psf files with a 15Å additional layer of watermolecules from the most outer portion of the protein (B R O O K Set al. 2009). The time step of the simulation was 1 fs. Periodicboundary condition was used. For equilibration, 5000 steps ofenergy minimization with conjugate gradient method has beenperformed initially, then the temperature was raised from abso-lute zero to 37° in one degree intervals using Langevin dynamicswith 1 ps of simulation at the temperature. Then, five cycles of10 ps simulation with 2000 steps of energy minimization wereperformed. Each simulation was performed for 20 additional ns(videos provided in Files S2-S5).

Data AvailabilityStrains are available upon request. Table S1 contains primersused for daf-2 sequencing. Table S2 contains a complete list ofthe dauer assay results by independent isolate. Table S3 con-tains dauer-associated mutations in pre-backcrossed MMP strains.Table S4 contains the CRISPR-Cas9 gRNA and dsDNA repair tem-plate sequences with associated protocols. Table S5 contains SIFT,Polyphen2, AlignGVGD, and PROVEAN results. Table S6 con-tains quality scores and metrics for the DAF-2 I-TASSER model(Figure 4). Table S7 contains PDB IDs with supporting data forthe human tyrosine kinase structural analysis. Figure S1 containsthe human to worm sequence alignment annotated with the MMPmutations. File S1 contains the complete sequence alignment file.Files S2-5 contain molecular dynamics simulations highlightingthe ATP movements in the tyrosine kinase binding pocket of nor-mal and mutant human INSR. High resolution videos are availableupon request. File S6 contains the human tyrosine kinase activesite alignment. Files S7-S13 contain the DAF-2 I-TASSER modelPDB files (Figure 4).

RESULTS

Recent advances in worm genetics and genomics have acceleratedthe ability to generate, map and characterize mutant alleles. We de-vised a pipeline to examine the extent to which these technologicaladvances could be applied to a well-characterized and evolution-arily conserved pathway to both gain a deeper understanding ofstructure-function relationships and to model human disease. Theinsulin-like signaling pathway was chosen for a detailed analysis.As shown in Figure 1, this pipeline was designed to use informa-tion derived from existing mutations in the human IIS as a startingpoint and to use C. elegans genetics to model mutations in con-served regions of components of this well-studied pathway. Thestrategy used bioinformatic tools coupled with allele selection andgene editing to examine a robust insulin-dependent phenotype inan intact organism, focusing on a high-throughput dauer assay.

DAF-2 MMP alleles as a proof-of-principle targetAlleles available from the Million Mutation Project (MMP) wereselected solely on the basis of survival at room temperature(T H O M P S O N et al. 2013). One goal of this study was to determineif MMP alleles would have value in further understanding the IISpathway. In man, defects in insulin signaling can present with awide range of phenotypic consequences for the affected patient.

Modeling this phenotypic spectrum may require the identificationof more subtle phenotypes than traditional screens allow.

To fully understand the scope of IIS pathway mutants avail-able in C. elegans, we interrogated WormBase (v245) for all knownalleles in evolutionarily conserved IIS pathway genes (H O W Eet al. 2016), including genetic mutants, deletion alleles, and non-synonymous single nucleotide polymorphisms (nsSNPs) identifiedby the MMP. As Figure 2A demonstrates, the MMP nsSNPs greatlyexpand the range of mutations available for the components of theIIS pathway. We focused this study on the sole C. elegans insulin-like receptor, DAF-2, for several reasons: 1) 40 new missense alleleswere present in the MMP collection allowing us to further assessDAF-2 structure-function; 2) reduction of DAF-2 activity results inan easily scored dauer phenotype in a dose-dependent manner; 3)25 mutants resulting in a dauer or more severe larval arrest pheno-type with a known genotype have previously been characterized;and 4) the protein is closely related to the human insulin receptor(INSR) which has been extensively characterized in patients withsevere insulin resistance and morbidity due to INSR mutations(G E M S et al. 1998; PAT E L et al. 2008).

The available daf-2 MMP nsSNP alleles were distributed acrossthe entire gene locus with hits affecting exons encoding all majordomains of the protein (Figures 3-4). There was an average of 8.7MMP mutations per kilobase (Kb) in daf-2 with hit frequencies sim-ilar in exons (10/Kb) and introns (8.4/Kb); a similar distributionwas observed across other IIS pathway genes (T H O M P S O N et al.2013). Complete loss of daf-2 function results in embryonic lethalityor early larval arrest (G E M S et al. 1998), thus none of the MMPalleles would be expected to be null on their own. It is possible thathomozygous daf-2 lethal alleles could be among the MMP collec-tion due to co-selection for second site suppressors of the lethality.Previous work has shown mutations in DAF-18/PTEN can act assuch a suppressor (O G G and R U V K U N 1998). In fact, three daf-2MMP strains we selected (VC40071, VC41001, and VC40691) alsohad nsSNPs in daf-18, which might allow the strains to survive theMMP room-temperature viability requirement even with a loss-of-function daf-2 mutation (Table S3). This highlights the value of thecomplete genome sequence and the need to genetically outcrossthese strains prior to further study (see below and Materials andMethods).

To determine the distribution of nsSNP alleles among evolution-arily conserved and non-conserved domains of DAF-2, we alignedrelated protein sequences both from within and between genera(File S1 and Figure S1). Six Caenorhabditid species with sequencedgenomes were included in this analysis: C. elegans, C. remanei, C.briggsae, C. brenneri, C. sp5, and C. tropicalis. We also includedgenera that had well-annotated genomic, transcriptomic and pro-teomic evidence for an insulin-like receptor (human, chimp, mouse,rat, frog, zebrafish, and fly). In order to correct for variable-sizedinsertions and deletions of non-conserved amino acids in DAF-2relative to related receptors, T-COFFEE multiple-sequence align-ments were performed with anchors at known functional sequencemotifs surrounding clearly identifiable domains, including the in-sulin binding domain, fibronectin type III repeats, trans-membranedomain, and kinase domain (M O R G E N S T E R N et al. 2006; TA LYet al. 2011). This approach gave improved sequence alignmentsthat allowed us to determine the degree of amino acid conservationboth between and within genera (Table 1, Figure S1, and File S1).

As expected, the majority of MMP nsSNPs (28/40) were locatedin regions that were not evolutionarily conserved in contrast toprevious phenotypic screens that isolated mutations only in highlyconserved domains of the DAF-2 receptor (Figure 2B). Among

Volume X November 2016 | Worm Model of Insulin Receptoropathy | 3

n Table 1 Conservation and phenotypic consequences of novel DAF-2 mutations.

Genotype Change Cons Effect Prediction Hid %D STDEV SE N I.I.

WT 2.0 0.3 0.3 6953 1gk402833 R10Q no Charge N.A. Yes 57.4 44.7 22.4 2328 4gk169977 L24F no minor N.A. no 0.5 0.3 0.2 2749 3gk407256 R49G no Major N.A. no 0.0 0.0 0.0 665 3gk680522 P155L Yes* Major N.A. Yes 60.3 36.4 21.0 1703 3gk819989 V279I no minor N.A. no 0.6 0.4 0.3 323 2gk169945 S341P Yes* Major N.A. Yes 44.3 10.4 6.0 2397 3gk617200 G354S no Major N.A. no 0.1 0.1 0.1 1666 3gk595034 R375C no Charge N.A. Yes 86.4 6.3 3.6 1999 3gk169944 D467N no Charge N.A. no 0.9 2.2 0.9 2234 7

CB120 P470L I. Major LoF Yes 82.7 18.2 10.5 765 3JT6 P470L I. Major LoF Yes 90.0 6.6 3.3 1485 4

gk745372 E482K no Ch. Rev N.A. no 0.0 0.0 0.0 515 4gk899879 A508T I. Polarity LoF No 2.5 2.1 1.2 14 507 3gk379607 H543Y no Charge N.A. no 0.3 1.6 0.8 2436 4gk409799 N610K no Charge N.A. no 0.3 0.3 0.2 2167 4gk461841 V653M Yes† minor N.A. no 1.7 2.4 1.4 1600 3gk748865 A746T Yes‡ Polarity N.A. Yes 98.6 1.2 0.7 1218 3gk808587 T775M no Polarity N.A. no 0.0 0.0 0.0 1009 3gk169932 M868V no minor N.A. no 0.0 0.0 0.0 3728 3gk351031 D902N no Charge N.A. no 1.1 1.0 0.6 1303 3gk381236 A906V Yes§ minor N.A. no 0.3 0.7 0.4 3124 3gk437687 N937S no minor N.A. no 0.3 0.6 0.4 618 3gk845453 S940N no minor N.A. no 0.2 0.3 0.2 464 3gk452089 G958R no Charge N.A. no 0.0 0.0 0.0 332 3gk922267 R1002Q no Charge N.A. no 0.2 0.5 0.4 1711 2gk660898 E1008K no Ch. Rev N.A. no 0.0 0.0 0.0 980 3gk704420 E1005K no Ch. Rev N.A. no 0.6 1.0 0.7 1464 2gk169926 A1015V no minor N.A. Yes 62.6 53.1 30.7 1247 3gk828902 I1029M no minor N.A. no 0.0 0.0 0.0 1004 3gk373225 P1098S no Major N.A. no 0.4 0.7 0.4 1604 3gk606297 L1189F Yes** minor N.A. no 0.0 0.0 0.0 595 3gk574441 S1290L no Polarity N.A. no 0.0 0.0 0.0 473 3gk577128 E1380K I. Ch. Rev LoF Yes 34.7 44.3 25.6 671 3gk169915 A1391T I. Polarity LoF No 0.2 0.3 0.2 580 4gk169914 R1398Q no Charge N.A. no 0.0 0.0 0.0 1900 2gk390525 K1614E no Charge N.A. no 0.1 0.1 0.1 3561 3gk735782 S1636N no minor N.A. no 0.5 0.8 0.6 1869 2gk764453 D1653N no Charge N.A. no 0.1 0.1 0.1 3634 3gk402261 A1729V no minor N.A. Yes 12.9 18.7 9.4 1693 4gk913536 G1746D no Charge N.A. no 3.1 4.1 2.4 926 3gk654266 R1779Q no Charge N.A. no 1.0 2.3 1.3 1116 3gk559119 D1797N no Charge N.A. no 1.4 0.2 0.1 2430 2

For each allele, the genotype, residue change (Change), degree of conservation (Cons), amino acid property change (Effect), and predictedphenotypic effect (Prediction) are indicated. In addition, high-temperature induction of dauer (Hid) phenotype, percent dauer (%D), standarddeviation (STDEV), standard error (SE), number of worms assayed (N), and number of independent isolates (I.I.) are shown. Residue numberingrefers to the preproreceptor. *Substituting residue is found at the corresponding location in a human insulin-like receptor. †Substituting residuemaintains conserved amino acid charge. ‡Identical residue found in 22/27 aligned sequences (File S1). §Conserved charge change. **Substitutingresidue maintains hydrophobicity of the transmembrane domain. WT, Wild Type; I., identical; Ch. Rev., charge reversal. Colors are shown asdefined in Figure 3.

4 | D. A. Bulger et al.

those SNPs in highly conserved domains, four altered residuesthat were identical when comparing human and C. elegans insulinreceptors. These four allowed us to apply multiple algorithms,including PolyPhen2, SIFT, Align GVGD, and PROVEAN (TableS5), to predict the potential functional consequences of these substi-tutions (K U M A R et al. 2009; A D Z H U B E I et al. 2010; H I C K S et al.2011; C H O I et al. 2012). Each of these four mutants, two withinthe insulin binding domain ((DAF-2(P470L), DAF-2(A508T)) andtwo within the tyrosine kinase domain (DAF-2(E1380K), DAF-2(A1391T)), was predicted by multiple analytic programs to bedeleterious, suggesting they should result in a phenotype. Forthe remaining 8 MMP alleles that altered conserved non-identicalresidues, the anchored sequence alignment and functional motifrequirements of each domain were assessed. Three of these mu-tants (DAF-2(V653M), DAF-2(A906V), and DAF-2(L1189F)) werepredicted not to affect function because the missense mutationretained conserved hydrophobicity at the residue position (Ta-bles 1-2; asterisks). Similarly, three mutants (DAF-2(P155L), DAF-2(S341P), and DAF-2(A746T)) resulted in moderately conservativesubstitutions that might not be expected to affect function (Table 1;daggers). We concluded that the MMP daf-2 allele collection pro-vided a useful mix of amino acid substitutions, including severalchanges in conserved residues that had a high likelihood of af-fecting function and, therefore, would likely have a dauer-relatedphenotype.

Multiple MMP daf-2 alleles have a high temperature induction ofdauer (Hid) phenotype

The large burden of mutations in the original MMP strains, visi-ble in the genomic sequence data, often resulted in poor growthand high lethality; the number of nsSNPs per strain for the 40daf-2 allele-containing strains studied ranged from 198 to 1,025(mean ± STDEV of 471.6 ± 187). Two genetic backcrosses wereperformed for each MMP strain using a genetic marker for thedaf-2 region (let-805st456/qC1 [dpy-19(e1259ts) glp-1(q339)]) and atleast two independent out-crossed isolates were selected for eachallele. The independent out-crossed isolates allowed us to controlfor the influence of residual background mutations in our assays,a potential confounding factor given some strains had up to fouradditional mutations in genes known to affect dauer formation(Table S3). With a few exceptions (Table S2), none of the indepen-dent isolates after backcrossing had noticeable phenotypes at roomtemperature.

We screened each of the out-crossed MMP allele isolates forDAF-2 function by scoring for a high temperature induction ofdauer (Hid) phenotype. High-temperature (27.2°) has previouslybeen shown to induce dauers at low frequency in wild type ani-mals and to enhance the dauer phenotype of weak DAF-2 loss-of-function alleles (A I L I O N and T H O M A S 2003). We also assayedtwo additional strains (CB120 and JT6) that were previously iden-tified in genetic screens for constitutive dauers (Daf-c), but forwhich the molecular lesion was unknown; sequencing the daf-2locus in each strain allowed us to identify the causative mutation(see below).

Of the 42 strains harboring nsSNPs assayed for a Hid pheno-type at 27.2°, four strains had a clear dauer phenotype (> 80%dauer with SE < 15), six strains had a variable dauer phenotype,and 32 strains had no observable dauer phenotype (< 3.5% dauerwith SE < 2.5) (Table 1). All four strains resulting in a clear dauerphenotype had mutations in the insulin binding domain (aminoacid residues 140-774). Two of these were novel MMP strains(DAF-2(R375C) and DAF-2(A746T)) and two were previously iden-

tified strains (CB120 and JT6) in which we identified the sameDAF-2(P470L) mutation. Two of these three Hid-associated substi-tutions affected highly conserved amino acid residues. Alanine746

was conserved between 22 of the 27 receptors we aligned withonly 5/11 Drosophila strains having a serine instead of an alanine.Proline470, altered in strains CB120 and JT6, was conserved in 25 ofthe 27 aligned species with only the two Branchiostoma species hav-ing a glutamate at that position. The less-conserved residue changethat also resulted in a clear dauer phenotype, DAF-2(R375C), wasonly conserved in four out of the six aligned Caenorhabditids (C.elegans, C. remanei, C. briggsae, and C. sp5) and 8/21 of the remain-ing aligned species with 10 of the other species having either alysine or a proline at this position. Thus, while strong evolutionaryconservation of a residue was a very good predictor for mutationsresulting in mutant phenotypes, substitutions in residues withweaker conservation also had functional consequences.

The six strains with variable dauer phenotypes affected aminoacid residues dispersed across five different domains of the pro-tein: DAF-2(R10Q) in the signal peptide, DAF-2(P155L) and DAF-2(S341P) in the insulin binding domain, DAF-2(A1015V) in thefibronectin type III domain, DAF-2(E1380K) in the tyrosine kinasedomain, and DAF-2(A1729V) in the C-terminal domain. The signalpeptide mutant, DAF-2(R10Q), was in a residue located after thefirst SL1 trans-splice site (K R A U S E and H I R S H 1987), the majortrans-spliced isoform, which is part of a potential (but experimen-tally unconfirmed) RXRR furin cleavage site. The remaining fivevariable Hid-associated strains similarly affected residues withonly moderate to low conservation, weakening our confidencein attributing the phenotype to the daf-2 nsSNP. The proline155

residue is conserved only among worms (Caenorhabditids and Bru-gia) with nonpolar amino acids (L, I, V, A) in the other alignedhigher eukaryotes. Because leucine is found at this position inhigher eukaryotes, the MMP DAF-2(P155L) mutation was not ex-pected to yield a dauer phenotype. The other insulin bindingdomain substitution (DAF-2(S341P)) affected a serine conservedonly in the two aligned Xenopus species, with higher eukaryotes(mouse, rat, chimp, man) having a threonine at that position. TheDAF-2(A1015V) substitution is in a residue that is neither con-served among worms nor the other aligned organisms. The DAF-2(E1380K) substitution affects a tyrosine kinase glutamate residuethat is conserved only among Caenorhabditids, at the same site asthe glutamine thought to anchor the E-helix to the F-helix in the hu-man INSR. Finally, the DAF-2(A1729V) change had a weak dauerphenotype with a standard deviation greater than the mean per-centage of dauers (12.9 ± 18.7%). This residue is only conservedamong Caenorhabditids and this region of the C-terminal domain isunique to this genus. Based on relatively weak sequence conser-vation of affected residues, we concluded that with these resultsalone we could not confirm that any of these substitutions causethe weak and variable dauer phenotypes.

The 32 MMP alleles that were indistinguishable from wild typecontrols in our Hid phenotype assay were mostly in non-conservedresidues distributed across the protein with two clear exceptions:DAF-2(A508T) in the insulin binding domain and DAF-2(A1391T)in the tyrosine kinase domain. DAF-2(A1391T) was especiallysurprising since it was in a conserved catalytic loop amino acidresidue found to be mutated in several patients with severe insulinresistance (C A M A et al. 1993; M O R E I R A et al. 2010). This mutationwill be discussed below in the context of our results from CRISPR-Cas9 genome edited strains.

Volume X November 2016 | Worm Model of Insulin Receptoropathy | 5

n Table 2 Validation of MMP phenotypes and modeling of conserved human disease alleles by CRISPR-Cas9.

Domain Change Conserved % Dauer Dauer EL Adult N

Binding P155L yes 0.2 1 88 450 539

Binding S341P yes* 0.0 0 82 942 1024

Binding A746T yes 99.3 1607 0 12 1619

FnIII A1015V no 41.5 235 16 315 566

Tyr K A1391T yes 0.0 0 0 374 374

Tyr K A1391V yes 98.7 548 0 7 555

C-term A1729V no 7.4 358 1213 3236 4807

The domains, residue changes, conservation, dauer phenotype and associated statistics for each of the engineered alleles are given. Residuenumbering refers to the preproreceptor. Vertical colored bars reflecting different protein domains are consistent throughout tables and figures asdefined in Figure 3. *Substituting residue is found at the corresponding location in a human insulin-like receptor. Dauer, number of dauer larvae;EL, number of embryonic lethal; Binding, ligand binding; FnIII, fibronectin Type III; Tyr K, tyrosine kinase; C-term, C-terminal.

Validating dauer phenotypes by CRISPR-Cas9 genome editing

To validate dauer phenotypes observed in the out-crossed strains,we carried out genome edits to mimic daf-2 MMP alleles in anotherwise isogenic, wild type genetic background using a dpy-10co-editing strategy (A R R I B E R E et al. 2014). Genome editing usingthe CRISPR-Cas9 system has become robust for many, but notall, target sites in the C. elegans genome (PA I X et al. 2014). Weinitially selected DAF-2 alterations R10Q, P155L, S341P, A746T,A1015V, E1380K, and A1729V for validation; we were successfulin generating at least two independent strains with precise editsfor all of these except R10Q and E1380K (Table S4).

Strains harboring the successfully CRISPR-Cas9 edited alleleswere each assayed for a Hid phenotype and the results comparedto our genetic alleles (Table 2). As expected, the edited allelegenerating the insulin binding domain alteration A746T had astrong dauer phenotype, matching the MMP genetic results for thissubstitution. Two genomic edits resulting in substitutions (A1015Vand A1729V) had moderate Hid phenotypes, also consistent withthe MMP genetic alleles. Interestingly, the A1729V substitutionhad a consistently low percentage of dauer formation in boththe MMP out-crossed strains (12.9%) and CRISPR-Cas9 editedstrains (7.4%); we concluded this amino acid substitution wascausative. This is the first dauer-enhancing allele identified in theDAF-2 C-terminal domain about which little is known functionally.Finally, two DAF-2 substitutions (P155L and S341P) had no Hidphenotype, in contrast to our genetic results. Thus, analysis ofCRISPR-Cas9 edited alleles was essential in validating which ofthe MMP substitutions were causative for the dauer phenotype.Based on this analysis, 7 of the assayed strains (5 MMP, CB120,JT6) were validated as true Hid positives while 35 MMP allelesresulted in a phenotype indistinguishable from wild type in thehigh temperature-induced dauer assay.

C. elegans DAF-2 mutations faithfully model human disease alle-les

Of the human INSR mutations associated with disease, five corre-spond to conserved residues in DAF-2 for which mutations havepreviously been identified that affect dauer formation (Table 3).A single MMP mutation (DAF-2(A1391T)) we assayed matcheda known human INSR allele, but surprisingly resulted in no ob-servable Hid phenotype. This DAF-2 variant aligned with boththe INSR(A1135E) mutation found in a patient with Type A In-

sulin Resistance and the compound heterozygous INSR(A1135V;N878S) mutation found in a patient with Rabson-Mendenhall Syn-drome, a very severe form of insulin resistance. The lack of a Hidphenotype in this strain suggested that the C. elegans threoninesubstitution was tolerated despite the bioinformatic predictions tothe contrary. This residue was chosen for further study to deter-mine if mimicking the human disease alleles in C. elegans wouldprovide additional insight.

We generated three alanine1391 substitutions in C. elegans DAF-2 using CRISPR-Cas9 genomic editing and assayed each for aHid phenotype. One change (DAF-2(A1391T) replicated the MMPallele that had no Hid phenotype. The other two changes (DAF-2(A1391V) and (A1391E)) mimicked human INSR substitutionsresulting in disease. The DAF-2(A1391T) edited strains had noobservable dauer phenotype, reproducing the results from thecorresponding MMP strain confirming that this substitution istolerated in DAF-2. In contrast, the two other human disease sub-stitution mimics had Hid phenotypes, although both the severityand penetrance of the phenotype differed between them. TheDAF-2(A1391V) mutation resulted in 98.7% dauers at 27.2° and 0%dauer at 20°. Interestingly, the DAF-2(A1391E) edit could only bemaintained as a heterozygote; animals harboring this edit gave riseto 25% dauer progeny at 15°, 20°, and 25°, suggesting the homozy-gous mutation had a Daf-c phenotype (Table 4). We confirmedthis by genotyping individual progeny of these putative heterozy-gous animals and found that 10 out of 10 sequenced individualdauer progeny were homozygous for the mutation and none of30 sequenced non-dauer progeny were homozygous for the muta-tion (only wild-type or heterozygous). Thus, the DAF-2(A1391E)homozygotes were found to be constitutive dauer at all tempera-tures tested, mimicking the severe phenotype observed in humansheterozygous for this same substitution. Importantly, our DAF-2edits generated an allelic series for variants at this residue thatwere consistent with human disease severity: A1135T (tolerated),A1135V (moderate defect), and A1135E (severe defect).

We explored these daf-2 alanine1391 substitutions in more detailto understand the molecular basis of the lesions. Alanine1391 islocated at the center of the tyrosine kinase active site and is highlyconserved. Sequence alignment between the HRD and DFG motifs,which define a key 23 amino acid long region of the tyrosine kinaseactive site (W H E E L E R and YA R D E N 2015), demonstrated thatan alanine is found in 79 out of the 90 human tyrosine kinases wealigned; 9 of 90 human tyrosine kinases have a threonine at this site

6 | D. A. Bulger et al.

and only one (FAK2) has a valine (a possible loss-of-function vari-ant) (File S6). Thus, like our in vivo results in C. elegans, sequencealignments suggest threonine may be a tolerated substitution atthis position in humans.

To further understand the molecular consequences of sub-stitutions at human INSR alanine1135 (corresponding to DAF-2alanine1391), we turned to available structural data. Due to thenumerous human diseases caused by tyrosine kinase dysfunction,several human tyrosine kinase crystal structures were availableto explore the potential consequences of substitutions at this site(B E R M A N et al. 2002). In the crystal structures for seven of thenine human tyrosine kinases with a threonine at this position, thethreonine is stabilized through a hydrogen bond with a nearbyglutamate (Table S7). To see if this might also be the case for thehuman INSR(A1135T) variant, CHARMM molecular dynamicssimulation with periodic water solvation was performed for 20ns on the human INSR tyrosine kinase crystal structure with thewild-type alanine or an identical sequence with that residue sub-stituted with a threonine (videos in Files S5). We found that thewild type and A1135T variant INSR simulations resulted in no sig-nificant changes in the stability of the tyrosine kinase domain andactive site configuration. We extended this analysis to determinethe potential effects of the valine and glutamate substitutions atthis residue that are associated with human disease. Althoughthe valine change had no observable effects during simulation,the A1135E mutation significantly disrupted the active site sta-bility. The human INSR(A1135E) has previously been shown tohave processing defects in the ER preventing >95% of the receptorfrom reaching the cell surface (C A M A et al. 1993). Our simula-tion demonstrated that any INSR(A1135E) making it to the cellsurface would likely be compromised for function based on a veryunstable kinase active site.

DISCUSSION

An emerging spectrum of human insulin receptoropathiesOver 150 different alleles have already been described in the hu-man insulin receptor; this list is likely to expand greatly in thenear future owing to the ever-decreasing cost of acquiring genomesequence information and the many efforts world-wide in collect-ing large cohorts of sequenced genomes. A key challenge for themedical research community is to identify those alleles that areassociated with disease risk and tailoring individual therapy forthe best patient outcome.

Like many conserved pathways, dysfunction of insulin/IGF-1signaling (IIS) can result in a range of disease phenotypes depend-ing on the nature of the lesion. Thus, the phenotypic consequencesof INSR variants are graded from none to very severe, reflect-ing the extent to which signaling is compromised. It is generallyaccepted that INSR polymorphisms retaining greater than 50%receptor function do not result in human disease (K A D O WA K Iet al. 1988; TAY L O R et al. 1992; S E M P L E et al. 2011; S E M P L E2016; S E M P L E et al. 2016). However, available studies suggestthat reduction of INSR function to approximately the 25-50% rangeresults in Type A Insulin Resistance (TA-IR), in which patientspresent with hyperinsulinemia, acanthosis nigricans, polycysticovarian syndrome (PCOS), hirsuitism, and hyperandrogenism.Patients with less than 25% INSR function have the more severedisorders Rabson Mendenhall syndrome (RMS) and Donohue syn-drome (DS; formerly “leprechaunism”). Patients with RMS and DShave additional symptoms of linear growth retardation, impaireddevelopment of skeletal muscle and adipose tissue, hypertrichosis,and additional dysmorphic features. While patients with TA-IR

have normal lifespans, RMS and DS patients have a much worseprognosis typically living 3-17 years for RMS and less than 3 yearsfor DS. These clinical outcomes underscore the importance of fullyunderstanding the functional consequences of INSR mutations.

C. elegans as a model for insulin-like signaling

The worm has proven to be a genetically amenable model for exam-ining conserved biological pathways such as the insulin-signalingpathway, which is deregulated in many human diseases. Wormgenetics provided an early framework for understanding the or-der of this complex signaling cascade. We developed the pipelineshown in Figure 1 to determine if C. elegans genetic tools could 1)improve our understanding of the well-studied insulin-like recep-tor DAF-2 and 2) prove useful for the analysis of human metabolicdisease for which sequence polymorphisms were readily availablefrom exon-sequencing efforts. Using a positive selection screenfor daf-2 mutants, David Gems’ group has investigated the molec-ular basis of the different alleles resulting in a dauer-associatedphenotype. Their analysis resulted in the clustering of existentgenetic alleles into two classes (I and II). Furthermore, they devel-oped an online tool (RILM) that contains a sequence alignmentacross many species allowing them to note that the two Class Imutants daf-2(m577) and daf-2(m596) occur at residues with knownhuman insulin receptor disease-causing mutations, both of whichare thought to cause processing defects. Here we show numer-ous other Class II daf-2 mutants that also occur in known humandisease-causing residues (Table 3). While class distinction hasproven useful for the more severe Daf-c causing mutations, thisdistinction has not been made for the weaker high-temperature in-duction of dauer (Hid) mutants presented in this study. Our resultsdemonstrate that some human disease-associated allele mimicsin DAF-2 present with weaker phenotypes, highlighting the im-portance of the biological assay and its sensitivity in interpretingfunction.

Our study also tested the utility of the Million Mutation Project(MMP) resource in providing new insights into the highly con-served and well-studied insulin/IGF-1 signaling (IIS) pathway.Phenotyping 40 daf-2 MMP alleles using a relatively simple andquick high temperature-induced dauer assay, we identify 5 newdauer alleles and 35 tolerated DAF-2 substitutions (Table 1). Wefound that accurately predicting phenotypes for animals harboringeven severe amino acid substitutions (e.g. charge reversal) wasnot possible, with many non-conservative changes resulting in nophenotype whereas even conservative changes could lead to dauerphenotypes. Our study expands the regions of DAF-2 exploredby mutagenesis for functional consequences while revealing clearparallels to human disease alleles in the evolutionarily conservedinsulin/IGF-1 signaling (IIS) pathway.

Human disease alleles in the conserved INSR receptor are effec-tively modeled in C. elegans

One of the twice-backcrossed MMP daf-2 alleles lacking a Hidphenotype (gk169915, DAF-2(A1391T)) aligns with known humandisease alleles at the corresponding position in INSR (A1135V ina complex heterozygote; A1135E in a heterozygote) (C A M A et al.1993; M O R E I R A et al. 2010). The absence of a Hid phenotype in C.elegans for DAF-2(A1391T), that we confirmed by genomic editingin an otherwise wild type background, is surprising because ofthe change in charge for this conserved residue within the tyrosinekinase domain that is predicted to be more deleterious than theA to V change previously associated with human disease. Ourgenetic results, coupled with the observation that the alanine to

Volume X November 2016 | Worm Model of Insulin Receptoropathy | 7

HUMAN METABOLIC DISEASE

INSULIN SIGNALING

INSULIN RECEPTOR

DEFICIENCY

HUMAN-WORM

T-COFFEE ALIGNMENT

IDENTICAL DAF-2 RESIDUES

I-TASSER STRUCTURAL MODEL

CHARM MOLECULAR

DYNAMICS SIMULATION

CRISPR-CAS9 MODEL

DAUER (HID) PHENOTYPE

MMP ALLELESDAF-D/DAF-C

OUTCROSSINGOUTCROSSING

Figure 1 Strategy and analysis pipeline exploiting human insulin-receptoropathy-associated polymorphisms and linking them toemerging C. elegans genetic tools. Among the components of the insulin signaling pathway, the human insulin receptor was chosen for itshigh degree of sequence similarity with C. elegans DAF-2. A combination of existing dauer mutants (Daf-defective and Daf-constitutive), MillionMutation Project (MMP) alleles and CRISPR-Cas9 generated alleles were coupled with a recursive verification of a high-temperature inductionof dauer (Hid) phenotype to provide a detailed look at conserved regions of the human insulin receptor and the functional consequences ofconservative and non-conservative replacements of critical residues in the tyrosine kinase domain.

8 | D. A. Bulger et al.

0 10 20 30 40 50 60 70 80

DAF-2

IST-1

AGE-1

AAP-1

DAF-18

AKT-1

AKT-2

SGK-1

DAF-16

Other

Missense

MMP

Insulin

INSR

IRS-1

PI3K

PTEN

PDK-1

AKT

FOXO

Signaling Cascade

Human C. elegans# Alleles

Phenotypic Alleles MMP Alleles

Nucleus

Cytosol

Extracellular

Conserved

Non-conserved

A

B

27

5

28

12

Figure 2 The evolutionarily conserved insulin-like signaling pathway members, associated alleles, and their distributions. (A) The C.elegans and human insulin signaling pathways are highly conserved from the DAF-2/INSR insulin-like receptor all the way to the DAF-16/FOXOnuclear transcription factor (pathway adapted from (C H R I S T E N S E N et al. 2002)). Only the pathway proteins associated with human insulinresistance and/or C. elegans dauer phenotypes are shown in this simplified diagram. Dauer-associated alleles identified in previous geneticscreens are indicted in red and grey bars labeled “Missense” and “Other,” respectively. Million Mutation Project (MMP) alleles are shown by bluebars. (B) Distinct distributions of phenotypically selected versus MMP DAF-2 alleles among conserved and non-conserved residues. DAF-2alleles identified by phenotypic selection appear to be biased towards substitutions in evolutionarily conserved residues, as defined in Table 1.In contrast, MMP alleles that were selected only for viability at room temperature do not seem to exhibit this bias.

Volume X November 2016 | Worm Model of Insulin Receptoropathy | 9

n Table 3 Phenotypic consequences of human INSR and corresponding C. elegans DAF-2 mutations.

INSR ∆ Human Disorder DAF-2 ∆ Allele Phenotype Class

R252H Type A IRR437C m579 Daf-c 2

R252C Type A IR

G366R; V28A DS G547S m596 Daf-c 1

A1135V; N878S RMSA1391T gk169915 WT WT

A1391V CRISPR ts-Daf ?

A1135E Type A IR A1391E CRISPR Daf-c NC

R1174Q Type A IR (7 cases) R1430Q sa223 Daf-c 2

P1178L Type A IR (2 cases) P1434S e1391 Daf-c 2

P1209A Type A IR P1465S e1370 Daf-c 2

The human INSR residue change (INSR ∆), associated human disorder, worm DAF-2 residue change (DAF-2 ∆), worm allele name (Allele),associated worm phenotype, and class as defined in (G E M S et al. 1998; PAT E L et al. 2008) are shown. Residue numbering refers to thepreproreceptor. NC, non-conditional Daf-c; WT, wild-type; ?, unknown; DS, Donohue Syndrome; RMS, Rabson-Mendenhall Syndrome. Colors asdescribed in Figure 3.

threonine change is present in 10% of the 90 human tyrosine kinasedomains we inspected (File S6), suggests it may similarly be atolerated change in the human INSR (M A N N I N G et al. 2002). Fur-ther analysis of this alanine1391 residue demonstrated our abilityto leverage the relatively quick genomic editing in C. elegans tomodel multiple alleles associated with human disease at the corre-sponding position in INSR. In fact, the edited substitutions in C.elegans generate an allelic series of phenotypic severity, matchingand further clarifying that observed in human patients.

The human A to V1135 INSR mutation has only been observedas one of the components of a compound heterozygous mutation:INSR(N851S; A1135V). As Rabson-Mendenhall Syndrome (RMS) isonly seen with either homozygous INSR mutations or compoundheterozygous mutations, it can be assumed that both mutationscause a significant loss-of-function. However, the degree of loss-of-function due to each allele is still unknown. The A to E substitutionat the same residue has already been shown in cell culture tocause mis-folding of the protein in the endoplasmic reticulum (ER)resulting in only a small fraction of this protein reaching the plasmamembrane to interact with ATP (C A M A et al. 1993). Whether the Ato V substitution causes a similar effect is unknown. Notably, the Ato T1135 INSR mutation has not yet been observed in humans eventhough 10% of human tyrosine kinases have a threonine at thisposition. Our results in C. elegans have demonstrated that thesethree substitutions constitute an allelic series, strongly suggestingthat human disease severity results from an analogous gradedeffect of these changes on INSR function.

The available crystal structures of human tyrosine kinases andmolecular dynamics simulations provide an approach to under-standing how mutations may induce graded effects on tyrosinekinase function (Table S7). By replacing the A1135 INSR residuewith a threonine, the approximately 3.0 Å distance between thesubstituted threonine’s hydroxyl group and the carboxylic acidgroup of E1201 suggested a potential hydrogen bond stabilizing thetwo charged residues in a relatively otherwise hydrophobic pocketof the protein. In the 10% of human tyrosine kinases normallycontaining a threonine corresponding to the A1135 INSR residue,a hydrogen bond can be seen in seven crystal structures providingfurther support for the INSR hydrogen bond explanation for how

the threonine replacement could occur without significantly alter-ing the stability of the protein. CHARMM molecular dynamicssimulations of all three missense mutations in the human INSRconfirmed the T1135-E1201 hydrogen bond and demonstrated thatthe A to T and A to V changes retained active site stability whilethe A to E change did not (videos in Files S2-S5). Thus, analysisand modeling of these three substitutions provides an explanationfor the genetic results observed in C. elegans and the degree ofdisease severity in humans.

Mis-folding of the INSR in the ER has been shown in humans tomediate many of the loss-of-function mutations (TAY L O R 1992).It would be interesting to directly address DAF-2(A1391E) proteinfolding, trafficking, and membrane localization, however, suitablereagents for these type of experiments have yet to be developed inC. elegans. Augmenting the facile genetics of the worm system within vivo DAF-2 cell biology will greatly advance our understand-ing the molecular mechanisms of dysfunction and add anotherdimension to modeling human INSR disease alleles.

Residue conservation predicts phenotypic consequences formutations

The MMP resource allows us to address the question of whetherpast phenotypic selections biased our collection of mutants affect-ing DAF-2 function. It did not; mutations in highly conservedresidues are very likely to have phenotypic consequence. Fivepreviously identified genetic alleles resulting in dauer formationdefects in C. elegans result from substitutions in residues conservedin human INSR; our current study adds another two to the list(Table 3).

Changes to evolutionarily constrained residues in the insulin-like receptor DAF-2 have phenotypic consequences in C. elegansthat correlate well with disease severity in humans with corre-sponding alterations. Of the five previously identified geneticreduction-of-function alleles in daf-2, each alters an amino acidresidue that is identical to either 27/27 or 26/27 related receptorsfrom multiple species that we aligned. The classic daf-2(e1370)allele encodes DAF-2(P1465S) that aligns with the INSR(P1209A)mutation in a patient with Type A Insulin Resistance (TA IR); thisproline residue is on the αF-helix of the tyrosine kinase domain.

10 | D. A. Bulger et al.

21

34

5

678

9

10

11

12

13

14

15

16

17

Exons Domains Screens MMP

*

*

*

*

*

*

*

*

SP

L1

CR

L2

Fn0

Fn1

Fn1

Fn2

TK

CT

Signal Peptide

Ligand Binding

Cysteine Rich

Ligand Binding

Fibronectin-like

Fibronectin-like

Fibronectin-like

TransmembraneJuxtamembrane

Tyrosine Kinase

C-terminus

Figure 3 Distribution of the position of phenotypic and MMP alleles in the exons and protein domains of DAF-2. The C. elegans daf-2exons (introns have been excluded) are shown at left in black and grey with the corresponding protein domains in color on the right. Schematicstyle adapted from (D E M E Y T S and W H I T TA K E R 2002). Mutations from previous phenotypic screens can be seen with missense mutationsin red and nonsense mutations in black; MMP alleles can be seen in blue on the right. Asterisks indicate which of the MMP alleles were foundto have novel dauer phenotypes in this study. While the exons and protein domains are shown to scale, the positions of the lines indicatingindividual mutations are approximate.

Volume X November 2016 | Worm Model of Insulin Receptoropathy | 11

Nonsense

MissenseMMPCysteine SS

Plasma Membrane

10 Å

FnIIICR

L2

L1α-N

β-N

β-C

α-C

R10Q

A508T

P155L

R375C

E1380K

A1391TA1391VA1391E

A1729V

TK

CT

Figure 4 Novel DAF-2 structural model highlighting known mutations including newly characterized MMP alleles and their phenotypicconsequences. A structural model of the DAF-2 insulin-like receptor is shown as a monomer based on domain-by-domain modeling using theonline I-TASSER server (http://zhanglab.ccmb.med.umich.edu/I-TASSER/) and I-TASSER 2.1 (Table S6 and Files S7-S13) (Z H A N G 2008).Missense and nonsense mutations from phenotypic screens and MMP mutations are shown as indicated in the legend. The domain labels havebeen color-coded to correspond to the protein domains in Figure 3. The predicted signal peptide sequence is represented by an N-terminaldotted line with the cleavage site indicated at residue 143 (scissors). The position and number of Fibronectin Type III-like domains is uncertaindue to an additional 75 amino acids that are present in DAF-2 when compared to human INSR (Figure 5). The putative furin-like cleavage siteis at residue 930 (scissors). The receptor alpha and beta strands resulting from furin cleavage are each labeled at their N- and C-terminal ends(N-α, C-α, N-β, C-β). The position and residue changes of MMP alleles resulting in a dauer phenotype are highlighted with black arrows andtext. Only MMP mutations with a dauer phenotype and the three CRISPR-Cas9 mutations at A1391 are identified. Residue numbering refers tothe preproreceptor. This method of displaying mutations was adapted from (PAT E L et al. 2008); the structure shown is a novel model of DAF-2based on I-TASSER structural modeling.

12 | D. A. Bulger et al.

n Table 4 Genomic edits confirming an MMP allele and modeling a human allelic series at the conserved residue INSR(A1135).

DAF-2 ∆1391 Dauer Assay Result 15° 20° 27°

A→T no phenotype N.D. 0 % 0 %

A→V temperature sensitive dauer N.D. 0 % 99 %

A→E constitutive dauer 100 % 100 % 100 %

CRISPR-Cas9 was used to alter DAF-2(A1391) to the indicated residues and the results are shown for dauer assays performed at indicatedtemperatures. As observed with human INSR(A1135) mutations, these alteration result in a graded range of phenotypes from no observablephenotype to constitutive dauer formation at all tested temperatures. N.D., not determined. Numbering refers to the preproreceptor. Color asdescribed in Figure 3.

The daf-2(e1391) allele encodes DAF-2(P1434S) that aligns with theINSR(P1178L) mutation located in the middle of the activation seg-ment of the tyrosine kinase domain and is commonly mutated inpatients with TA IR. The daf-2(sa223) allele encodes DAF-2(R1430Q)that perfectly matches the INSR(R1174Q) amino acid change foundin seven patients with TA IR. The daf-2(m579) allele encodes DAF-2(R437C) that perfectly matches a patient with TA IR and IR-ANwith an INSR(R252C) mutation and aligns with the INSR(R252H)mutation in another patient with TA IR. This arginine is in theinsulin-binding domain immediately adjacent to a completely con-served cysteine that likely plays a role in folding and may par-ticipate in an intra-chain disulfide bond. The daf-2(m596) alleleencodes DAF-2(G547S) that aligns with a compound heterozygouspatient with INSR(G366R; V28A) mutations resulting in DonohueSyndrome (DS), the most severe form of human insulin resistance.Of the aligned receptors, this glycine is conserved in the insulinbinding domains of human, chimp, mouse, rat, frog, zebrafish,and five out of six of the Caenorhabditids (C. briggsae had a gluta-mate at this residue). Thus, the most highly conserved domains ofthe INSR and DAF-2 are functionally dependent on the primaryamino acid sequence and substitutions have a high likelihood ofresulting in dysfunction; a similar conclusion was reached by (N Gand H E N I K O F F 2006).

Our current results using the MMP resource further supportthese conclusions, but also reveal at least one major surprise. Mostof the DAF-2 substitutions we assayed for a Hid phenotype, asingle yet sensitive metric for function, were wild type and innon-conserved residues. Therefore, the large majority of allelesfollowing mutagenesis and selection only for viability have noobvious deleterious effect on DAF-2 function. As described above,one of these (DAF-2(A1391T)) was, nonetheless, very informativein suggesting a functionally tolerated change that is unlikely toresult in human disease if identified in INSRs. The three MMPalleles that did result in a Hid phenotype identify novel residuesin which alterations would be predicted to have functional conse-quences for the human INSR and have a high likelihood of beingassociated with human disease; we would expect disease severityto correspond to the strength of the Hid phenotype in C. elegans.

Novel genetic insight into the C-terminal domain of DAF-2One of the dauer-enhancing MMP alleles, DAF-2(A1729V), con-firmed by CRISPR-Cas9 genome editing is located in the C-terminal domain, a stretch of 658 amino acids lacking previousgenetic hits. Although present in all INSR-related receptors, the C-terminal domain of the protein displays little to no conservation inits amino acid sequence or its length (Figure 5). The human IGF-1receptor has previously been shown to act as a dependence recep-tor, with the release of pro-apoptotic peptide cleavage products ofthe C-terminal domain during an absence of growth factor signal-

ing (O ’ C O N N O R et al. 1997; H O N G O et al. 1998; L I U et al. 1998;G O L D S C H N E I D E R and M E H L E N 2010). In flies, the C-terminaldomain of InR has the properties of Insulin Receptor Substrateproteins (IRSs) that in humans are encoded by four distinct genes(P O LT I L O V E et al. 2000). Fusion of the intracellular domains ofthe Drosophila InR protein to the human INSR ectodomain rescuesthe loss of INSR signaling in a mammalian cell line lacking allfour human IRS proteins. Interestingly, knockout of the sole flyortholog of IRSs, Chico, results in a completely different non-lethalphenotype than the InR knockout in flies. Therefore, while theC-terminal domain of InR in Drosophila can substitute functionallyfor absent human IRS proteins in tissue culture systems, theremay be additional in vivo functions of this domain in flies that areunrelated to canonical IIS.

Results from Drosophila may be relevant to understanding a rolefor the C. elegans DAF-2 C-terminal domain. DAF-2 and fly InRshare many characteristics, including being the sole insulin-like re-ceptor and both possessing a long C-terminal domain with numer-ous tyrosine residues (Figure 5). Like flies, C. elegans has a singleortholog of IRSs, IST-1. Mutations in ist-1 can enhance the dauerphenotype, but only when combined with daf-2 loss-of-functionmutations (W O L K O W et al. 2002), suggesting that IST-1, like Chicoin flies, is not essential for signaling through the IIS pathway. Thus,it seems plausible that the worm might share yet another featurewith the fly: the ability to signal directly through the C-terminal do-main instead of requiring recruitment of IRS proteins to the plasmamembrane. Our characterization of a weak dauer-enhancing MMPallele that results in DAF-2(A1729V) demonstrates for the first timethat the C-terminal domain is required for proper IIS signaling.

SummaryIn this study, insulin-like receptor signaling is interrogated bycharacterizing a rich array of novel C. elegans mutations, makingcomparisons to known human disease-causing insulin receptoralleles, and using whole animal in vivo functional assays of allelevariants. Capitalizing on the strengths of the worm system, includ-ing forward genetics, existing defined gene alleles, and relativelyeasy genomic editing by CRISPR-Cas9, we were able to quicklyphenotype 40 novel alleles while also creating and assaying hu-man allele mimics. Our results were a mix of the expected andunexpected, highlighting the importance of testing allele func-tion in the context of a whole animal system in which receptorprocessing and function are integrated with the entire metabolicprogram. Our experiences with the insulin-like signaling pathwayin C. elegans demonstrates the power of this model system to betterunderstand the possible organismal-level consequence of muta-tions in conserved physiological processes. Facile CRISPR-Cas9genomic editing makes feasible the assessment of any suitablehuman allele suspected of disrupting gene function in a model

Volume X November 2016 | Worm Model of Insulin Receptoropathy | 13

Worm DAF-2

Fly InR

Human IGF1R

Human INSR

NPXYXXM

SP

SP

L1

L1

L1

L1

CR

CR

CR

CR

L2

L2

L2

L2

FnIIIα FnIIIβ

FnIIIα FnIIIβ

FnIIIα FnIIIβ

FnIIIα FnIIIβ

TK

TK

TK

TK

CT

CT

CT

Figure 5 Conserved and divergent domains of four insulin-like preproreceptors. Protein schematics were drawn to scale relative to thenumber of amino acids in each domain. The worm DAF-2 domains were mapped onto the other receptors using anchored T-COFFEE align-ment (File S1). The Fly InR C-terminus contains multiple NPXYXXM motifs, which can bind both IRS-like proteins and PI3K (P O LT I L OV E et al.2000). The Human IGF1 ‘dependence receptor’ has a pro-apoptotic C-terminus (skull and cross-bones) (O ’ C O N N O R et al. 1997; H O N G O et al.1998; L I U et al. 1998; G O L D S C H N E I D E R and M E H L E N 2010). Unless indicated otherwise, domain designations are as shown in Figure 3.

system for which a robust assay exists or can be developed. Thisopens new possibilities for interpreting the increasing volumes ofhuman genome sequence data for identify mutations contributingto human diseases.

ACKNOWLEDGMENTS

We thank Phillip Gorden and Michelle Bond for useful discus-sions and comments, Andy Golden for CRISPR-Cas9 advice andreagents, Fred Dyda for assistance with structural models and in-sights, Thomas Brodigan for technical help throughout the project,and Wormbase (http://www.wormbase.org/). Some strains wereprovided by the CGC, which is funded by NIH Office of Re-search Infrastructure Programs (P40 OD010440). This work uti-lized the computational resources of the NIH HPC Biowulf cluster(http://hpc.nih.gov). This work was supported by the IntramuralResearch Program of the NIH, National Institute of Diabetes andDigestive and Kidney Diseases. DAB was supported by the NIHOxford-Cambridge Scholars Program. RKS was supported by theWellcome Trust [grant number WT098498] and the United King-dom National Institute for Health Research (NIHR) CambridgeBiomedical Research Centre.

LITERATURE CITED

Adzhubei, I. A., S. Schmidt, L. Peshkin, V. E. Ramensky, A. Gerasi-mova, et al., 2010 A method and server for predicting damagingmissense mutations. Nat Methods 7: 248–9.

Ailion, M., and J. H. Thomas, 2003 Isolation and characteriza-tion of high-temperature-induced dauer formation mutants inCaenorhabditis elegans. Genetics 165: 127–44.

Arribere, J. A., R. T. Bell, B. X. Fu, K. L. Artiles, P. S. Hartman, et al.,2014 Efficient marker-free recovery of custom genetic modifica-tions with CRISPR/Cas9 in Caenorhabditis elegans. Genetics 198:837–46.

Berman, H. M., T. Battistuz, T. N. Bhat, W. F. Bluhm, P. E. Bourne,et al., 2002 The protein data bank. Acta Crystallogr D Biol Crys-tallogr 58: 899–907.

Brenner, S., 1974 The genetics of Caenorhabditis elegans. Genetics 77:71–94.

Brooks, B. R., r. Brooks, C. L., J. Mackerell, A. D., L. Nilsson, R. J.Petrella, et al., 2009 CHARMM: the biomolecular simulationprogram. J Comput Chem 30: 1545–614.

Cama, A., M. de la Luz Sierra, M. J. Quon, L. Ottini, P. Gorden,et al., 1993 Substitution of glutamic acid for alanine 1135 inthe putative "catalytic loop" of the tyrosine kinase domain ofthe human insulin receptor. a mutation that impairs proteolyticprocessing into subunits and inhibits receptor tyrosine kinaseactivity. J Biol Chem 268: 8060–9.

Choi, Y., G. E. Sims, S. Murphy, J. R. Miller, and A. P. Chan, 2012Predicting the functional effect of amino acid substitutions andindels. PLoS ONE 7: e46688.

Christensen, K., T. E. Johnson, and J. W. Vaupel, 2002 The questfor genetic determinants of human longevity: challenges andinsights. Nat Rev Genet 7: 436–48.

De Meyts, P., and J. Whittaker, 2002 Structural biology of insulinand IGF1 receptors: implications for drug design. Nat Rev DrugDiscov 1: 769–83.

Dickinson, D. J., J. D. Ward, D. J. Reiner, and B. Goldstein, 2013 En-gineering the Caenorhabditis elegans genome using Cas9-triggeredhomologous recombination. Nat Methods 10: 1028–34.

Gems, D., A. J. Sutton, M. L. Sundermeyer, P. S. Albert, K. V. King,et al., 1998 Two pleiotropic classes of daf-2 mutation affect larvalarrest, adult behavior, reproduction and longevity in Caenorhab-ditis elegans. Genetics 150: 129–55.

Goldschneider, D., and P. Mehlen, 2010 Dependence receptors: anew paradigm in cell signaling and cancer therapy. Oncogene29: 1865–82.

Gottlieb, S., and G. Ruvkun, 1994 daf-2, daf-16 and daf-23: geneti-cally interacting genes controlling dauer formation in Caenorhab-ditis elegans. Genetics 137: 107–20.

Hicks, S., D. A. Wheeler, S. E. Plon, and M. Kimmel, 2011 Predic-tion of missense mutation functionality depends on both thealgorithm and sequence alignment employed. Human Mutation

14 | D. A. Bulger et al.

32: 661–668.Hongo, A., G. Yumet, M. Resnicoff, G. Romano, R. O’Connor, et al.,

1998 Inhibition of tumorigenesis and induction of apoptosis inhuman tumor cells by the stable expression of a myristylatedCOOH terminus of the insulin-like growth factor I receptor.Cancer Res 58: 2477–84.

Howe, K. L., B. J. Bolt, S. Cain, J. Chan, W. J. Chen, et al., 2016 Worm-Base 2016: expanding to enable helminth genomic research. Nu-cleic Acids Res 44: D774–80.

Hu, P. J., 2007 Dauer. WormBook : 1–19.Kadowaki, T., C. L. Bevins, A. Cama, K. Ojamaa, B. Marcus-

Samuels, et al., 1988 Two mutant alleles of the insulin receptorgene in a patient with extreme insulin resistance. Science 240:787–90.

Kimura, K. D., H. A. Tissenbaum, Y. Liu, and G. Ruvkun, 1997daf-2, an insulin receptor-like gene that regulates longevity anddiapause in Caenorhabditis elegans. Science 277: 942–6.

Krause, M., and D. Hirsh, 1987 A trans-spliced leader sequence onactin mRNA in C. elegans. Cell 49: 753–61.

Kumar, P., S. Henikoff, and P. C. Ng, 2009 Predicting the effects ofcoding non-synonymous variants on protein function using theSIFT algorithm. Nat Protoc 4: 1073–81.

Lee, H., M. K. Choi, D. Lee, H. S. Kim, H. Hwang, et al., 2011Nictation, a dispersal behavior of the nematode Caenorhabditiselegans, is regulated by IL2 neurons. Nat Neurosci 15: 107–12.

Lindhurst, M. J., V. E. Parker, F. Payne, J. C. Sapp, S. Rudge, et al.,2012 Mosaic overgrowth with fibroadipose hyperplasia is causedby somatic activating mutations in PIK3CA. Nat Genet 44: 928–33.

Liu, Y., S. Lehar, C. Corvi, G. Payne, and R. O’Connor, 1998 Expres-sion of the insulin-like growth factor I receptor C terminus as amyristylated protein leads to induction of apoptosis in tumorcells. Cancer Res 58: 570–6.

Malone, E. A., and J. H. Thomas, 1994 A screen for nonconditionaldauer-constitutive mutations in Caenorhabditis elegans. Genetics136: 879–86.

Manning, G., D. B. Whyte, R. Martinez, T. Hunter, and S. Su-darsanam, 2002 The protein kinase complement of the humangenome. Science 298: 1912–34.

Moreira, R. O., R. L. Zagury, T. S. Nascimento, and L. Zagury,2010 Multidrug therapy in a patient with Rabson-Mendenhallsyndrome. Diabetologia 53: 2454–5.

Morgenstern, B., S. J. Prohaska, D. Pohler, and P. F. Stadler, 2006Multiple sequence alignment with user-defined anchor points.Algorithms Mol Biol 1: 6.

Ng, P. C., and S. Henikoff, 2006 Predicting the effects of amino acidsubstitutions on protein function. Annu Rev Genomics HumGenet 7: 61–80.

O’Connor, R., A. Kauffmann-Zeh, Y. Liu, S. Lehar, G. I. Evan, et al.,1997 Identification of domains of the insulin-like growth factorI receptor that are required for protection from apoptosis. MolCell Biol 17: 427–35.

Ogg, S., and G. Ruvkun, 1998 The C. elegans PTEN homolog, DAF-18, acts in the insulin receptor-like metabolic signaling pathway.Molecular Cell 2: 887–893.

Paix, A., Y. Wang, H. E. Smith, C. Y. Lee, D. Calidas, et al., 2014Scalable and versatile genome editing using linear DNAs withmicrohomology to Cas9 sites in Caenorhabditis elegans. Genetics198: 1347–56.

Patel, D. S., A. Garza-Garcia, M. Nanji, J. J. McElwee, D. Ackerman,et al., 2008 Clustering of genetically defined allele classes in theCaenorhabditis elegans DAF-2 insulin/IGF-1 receptor. Genetics

178: 931–46.Poltilove, R. M., A. R. Jacobs, C. R. Haft, P. Xu, and S. I. Taylor,

2000 Characterization of Drosophila insulin receptor substrate. JBiol Chem 275: 23346–54.

Semple, R. K., 2016 EJE PRIZE 2015: How does insulin resistancearise, and how does it cause disease? human genetic lessons.Eur J Endocrinol 174: R209–23.

Semple, R. K., D. B. Savage, G. V. Brierley, and S. O’Rahilly, 2016Chapter 22 - syndromes of severe insulin resistance and/orlipodystrophy. In R. E. Weiss and S. Refetoff, editors, GeneticDiagnosis of Endocrine Disorders (Second edition). Academic Press,San Diego, second edition edition, 307 – 324.

Semple, R. K., D. B. Savage, E. K. Cochran, P. Gorden, andS. O’Rahilly, 2011 Genetic syndromes of severe insulin resistance.Endocr Rev 32: 498–514.

Taly, J. F., C. Magis, G. Bussotti, J. M. Chang, P. Di Tommaso, et al.,2011 Using the T-Coffee package to build multiple sequencealignments of protein, RNA, DNA sequences and 3D structures.Nat Protoc 6: 1669–82.

Taniguchi, C. M., B. Emanuelli, and C. R. Kahn, 2006 Critical nodesin signalling pathways: insights into insulin action. Nat RevMol Cell Biol 7: 85–96. 10.1038/nrm1837.

Taylor, S. I., 1992 Lilly lecture: molecular mechanisms of insulinresistance. lessons from patients with mutations in the insulin-receptor gene. Diabetes 41: 1473–90.

Taylor, S. I., A. Cama, D. Accili, F. Barbetti, M. J. Quon, et al., 1992Mutations in the insulin receptor gene. Endocr Rev 13: 566–95.

Thompson, O., M. Edgley, P. Strasbourger, S. Flibotte, B. Ewing,et al., 2013 The Million Mutation Project: A new approach togenetics in Caenorhabditis elegans. Genome Res .

Wheeler, D. L., and Y. Yarden, editors, 2015 Receptor Tyrosine Ki-nases: Structure, Functions and Role in Human. Humana PressSpringer, New York.

Wolkow, C. A., M. J. Munoz, D. L. Riddle, and G. Ruvkun, 2002Insulin receptor substrate and p55 orthologous adaptor proteinsfunction in the Caenorhabditis elegans daf-2/insulin-like signalingpathway. J Biol Chem 277: 49591–7.

Zhang, Y., 2008 I-TASSER server for protein 3D structure prediction.BMC Bioinformatics 9: 40. Zhang, Yang Research Support, Non-U.S. Gov’t England BMC Bioinformatics. 2008 Jan 23;9:40. doi:10.1186/1471-2105-9-40.

Volume X November 2016 | Worm Model of Insulin Receptoropathy | 15